[sheepdog] [PATCH 4/4] farm: use slice_{read, write} to read/write object

[Liu Yuan]

On Thu, Jul 18, 2013 at 04:35:20PM +0900, MORITA Kazutaka wrote:
> At Tue, 16 Jul 2013 17:30:21 +0800,
> Liu Yuan wrote:
> > 
> > For a test with 200M cluster with 2 copies (so roughly 100M data to backup),
> > I got the following resualt:
> > 
> >                    size  time
> > w/ slice (64K)  :  51M   2.037s
> > w/ slice (128K) :  53M   1.223s
> > w/ slice (256K) :  57M   1.216s
> > w/ slice (512K) :  61M   1.205s
> > w/o slice (4M)  :  97M   1.174s
> > 
> > This is why I choose 128K as slice size.
> > 
> > Signed-off-by: Liu Yuan <namei.unix at gmail.com>
> > ---
> >  collie/farm/farm.c |    4 ++--
> >  1 file changed, 2 insertions(+), 2 deletions(-)
> > 
> > diff --git a/collie/farm/farm.c b/collie/farm/farm.c
> > index 385397f..d4aff48 100644
> > --- a/collie/farm/farm.c
> > +++ b/collie/farm/farm.c
> > @@ -243,7 +243,7 @@ static void do_save_object(struct work *work)
> >  	if (sd_read_object(sw->entry.oid, buf, size, 0, true) < 0)
> >  		goto error;
> >  
> > -	if (sha1_file_write(buf, size, sw->entry.sha1) < 0)
> > +	if (slice_write(buf, size, sw->entry.sha1) < 0)
> >  		goto error;
> >  
> >  	free(buf);
> > @@ -337,7 +337,7 @@ static void do_load_object(struct work *work)
> >  
> >  	sw = container_of(work, struct snapshot_work, work);
> >  
> > -	buffer = sha1_file_read(sw->entry.sha1, &size);
> > +	buffer = slice_read(sw->entry.sha1, &size);
> >  
> >  	if (!buffer)
> >  		goto error;
> 
> Sorry, tests/functional/030 cannot be passed after merging this patch.

Well, I think it is something wrong with trim before sha1. I add my 1/4 patch
back and I passed the test. I noticed that load failure because it got a wrong
sha1. I guess it is some potential bug in trim-before-sha1.

I didn't find out the root cause yet, but I think would be an argument to merge
my 1/4 patch.

Thanks
Yuan